{"claims": [{"text": "The model achieved 95.63% segment-level accuracy on an independently held-out test set of 1,281 segments.", "quote_or_locator": "Results section: 'The optimized model achieved a multi-class segment-level accuracy of 95.63% on an independently held-out test partition of 1,281 segments'"}, {"text": "Internal five-fold cross-validation mean accuracy was 0.9846 ± 0.003.", "quote_or_locator": "Results section: 'with an internal five-fold cross-validation mean of 0.9846 ± 0.003'"}, {"text": "FTD-specific precision reached 0.9907.", "quote_or_locator": "Results section: 'FTD-specific precision reached 0.9907'"}, {"text": "The study included 88 participants: 36 with Alzheimer's disease (mean age 66.4 ± 7.9), 23 with frontotemporal dementia (mean age 63.6 ± 8.2), and 29 cognitively normal controls (mean age 67.9 ± 5.4).", "quote_or_locator": "Methodology section, Dataset subsection: 'this dataset contains resting-state eyes-closed (REST) recordings from a cohort of 88 participants, categorized into three groups: Alzheimer's Disease (36 AD patients; mean age 66.4 ± 7.9), Frontotemporal Dementia (23 FTD patients; mean age 63.6 ± 8.2), and Cognitively Normal controls (29 CN subjects; mean age 67.9 ± 5.4)'"}, {"text": "The initial feature pool was 1,014 dimensions, reduced to 200 via recursive feature elimination.", "quote_or_locator": "Methodology, Feature extraction section: 'The initial extraction phase yielded a comprehensive pool of 1,014 candidate features. To promote statistical parsimony and mitigate the risk of overfitting, the feature optimization protocol reduced the original feature space to the 200 most discriminative dimensions.'"}, {"text": "EEG data were acquired using a 19-channel system at 500 Hz sampling rate.", "quote_or_locator": "Methodology, Dataset section: 'All EEG data were acquired using a 19-channel system (Fp1, Fp2, F7, F3, Fz, F4, F8, T3, C3, Cz, C4, T4, T5, P3, Pz, P4, T6, O1, and O2) at a sampling rate of 500 Hz.'"}, {"text": "Resting-state beta-band hemispheric asymmetry was identified as the principal contributor to class separation via SHAP analysis.", "quote_or_locator": "Results section: 'SHAP analysis identified beta-band hemispheric asymmetry and alpha-band reactivity as the principal contributors to class separation.' and Discussion section: 'SHAP analysis identified Resting-state Beta Asymmetry (rest_asym_ch7-9_beta) as the highest-gain discriminative feature'"}, {"text": "The model achieved 97% sensitivity for Alzheimer's disease.", "quote_or_locator": "Results section, Table 3: 'Alzheimer's (AD) 97.0%' under Recall (Sens.) column"}, {"text": "AUC values were 0.998 for FTD, 0.995 for cognitive normal, and 0.993 for Alzheimer's disease.", "quote_or_locator": "Results section, ROC curve analysis: 'Frontotemporal dementia (FTD): AUC = 0.998' and 'Cognitive normal (CN): AUC = 0.995' and 'Alzheimer's disease (AD): AUC = 0.993'"}, {"text": "The weighted ensemble used 60% XGBoost and 40% Random Forest.", "quote_or_locator": "Methodology, Model architecture section: 'Pfinal = 0.6 × PXGBOOST + 0.4 × PRANDOMFOREST'"}, {"text": "The cohort was split subject-aware with 90% training and 10% test, yielding 11,620 training segments and 1,281 test segments.", "quote_or_locator": "Methodology, Data processing section: 'To prevent identity leakage, all 12,911 segments were grouped by participant ID and a subject-aware 90/10 block partition was applied. This allocated 11,620 segments for training and internal validation, and 1,281 segments to a strictly isolated holdout test set'"}, {"text": "Beta-band hemispheric asymmetry and alpha-band reactivity demonstrated 100% selection stability across the five cross-validation folds.", "quote_or_locator": "Discussion section, Model stability: 'While the exact composition of the 200-feature subspace varied slightly across the five folds, core physiological markers such as Beta Asymmetry and Alpha Reactivity demonstrated extremely high selection stability, appearing 100% of the fold-wise subsets'"}], "prompt_version": "p1.0", "verdicts": [{"claim": "The model achieved 95.63% segment-level accuracy on an independently held-out test set of 1,281 segments.", "verdict": "supported", "evidence": "The optimized model achieved a multi-class segment-level accuracy of 95.63% on an independently held-out test partition of 1,281 segments", "note": null}, {"claim": "Internal five-fold cross-validation mean accuracy was 0.9846 ± 0.003.", "verdict": "supported", "evidence": "with an internal five-fold cross-validation mean of 0.9846 ± 0.003", "note": null}, {"claim": "FTD-specific precision reached 0.9907.", "verdict": "supported", "evidence": "FTD-specific precision reached 0.9907", "note": null}, {"claim": "The study included 88 participants: 36 with Alzheimer's disease (mean age 66.4 ± 7.9), 23 with frontotemporal dementia (mean age 63.6 ± 8.2), and 29 cognitively normal controls (mean age 67.9 ± 5.4).", "verdict": "supported", "evidence": "this dataset contains resting-state eyes-closed (REST) recordings from a cohort of 88 participants, categorized into three groups: Alzheimer's Disease (36 AD patients; mean age 66.4 ± 7.9), Frontotemporal Dementia (23 FTD patients; mean age 63.6 ± 8.2), and Cognitively Normal controls (29 CN subjects; mean age 67.9 ± 5.4)", "note": null}, {"claim": "The initial feature pool was 1,014 dimensions, reduced to 200 via recursive feature elimination.", "verdict": "supported", "evidence": "The initial extraction phase yielded a comprehensive pool of 1,014 candidate features. To promote statistical parsimony and mitigate the risk of overfitting, the feature optimization protocol reduced the original feature space to the 200 most discriminative dimensions.", "note": null}, {"claim": "EEG data were acquired using a 19-channel system at 500 Hz sampling rate.", "verdict": "supported", "evidence": "All EEG data were acquired using a 19-channel system (Fp1, Fp2, F7, F3, Fz, F4, F8, T3, C3, Cz, C4, T4, T5, P3, Pz, P4, T6, O1, and O2) at a sampling rate of 500 Hz.", "note": null}, {"claim": "Resting-state beta-band hemispheric asymmetry was identified as the principal contributor to class separation via SHAP analysis.", "verdict": "distorted", "evidence": "SHAP analysis identified beta-band hemispheric asymmetry and alpha-band reactivity as the principal contributors to class separation.", "note": "The claim states only beta-band hemispheric asymmetry as the principal contributor, but the source indicates both beta-band hemispheric asymmetry AND alpha-band reactivity were identified as principal contributors. The claim omits alpha-band reactivity."}, {"claim": "The model achieved 97% sensitivity for Alzheimer's disease.", "verdict": "supported", "evidence": "The framework demonstrated high sensitivity for the Alzheimer's (AD) cohort (97.0%).", "note": null}, {"claim": "AUC values were 0.998 for FTD, 0.995 for cognitive normal, and 0.993 for Alzheimer's disease.", "verdict": "supported", "evidence": "Frontotemporal dementia (FTD): AUC = 0.998; Cognitive normal (CN): AUC = 0.995; Alzheimer's disease (AD): AUC = 0.993", "note": null}, {"claim": "The weighted ensemble used 60% XGBoost and 40% Random Forest.", "verdict": "supported", "evidence": "Pfinal = 0.6 × PXGBOOST + 0.4 × PRANDOMFOREST", "note": null}, {"claim": "The cohort was split subject-aware with 90% training and 10% test, yielding 11,620 training segments and 1,281 test segments.", "verdict": "supported", "evidence": "To prevent identity leakage, all 12,911 segments were grouped by participant ID and a subject-aware 90/10 block partition was applied. This allocated 11,620 segments for training and internal validation, and 1,281 segments to a strictly isolated holdout test set", "note": null}, {"claim": "Beta-band hemispheric asymmetry and alpha-band reactivity demonstrated 100% selection stability across the five cross-validation folds.", "verdict": "supported", "evidence": "While the exact composition of the 200-feature subspace varied slightly across the five folds, core physiological markers such as Beta Asymmetry and Alpha Reactivity demonstrated extremely high selection stability, appearing 100% of the fold-wise subsets", "note": null}]}